Claiming to monitor more than six times as many search terms as a standard lexicon, SteelEye’s product has been crafted to meet tightening regulations and growing volumes of communications channels.

Historically, financial institutions have relied on technology that scans internal communications for specific words and sequences of words to detect potential misconduct or market abuse, but its limited search capacity and irrelevant matches result in large number of false positives.

The lexicon developed by SteelEye covers tens of thousands of search terms to capture all linguistic variations, colloquialisms, ‘text-speak’, abbreviations, and even typos. Further, the lexicon also uses AI to filter for context and to remove false positives.

Despite many firms establishing policies that ban the use of alternative communications channels, like Whatsapp for instance, traders are still able to pursue these avenues to communicate. The data analytics firms grapples with this by flagging the context of a message and then the intent to take an illicit conversation to an unmonitored channel.

It does so by flagging a comment such as “Let’s talk on WhatApp” or “I’ll text you” - phrases which wouldn’t be recognised by conventional tools as ‘Whatsapp’ isn’t typically included in legacy lexicon, nor is the ability to discern the abbreviated ‘I’ll’ from ‘I will.’

Matt Storey, chief product officer at SteelEye says: “One of the main reasons legacy lexicons trigger such vast volumes of false positives is because they don’t consider the context in which a communication takes place.”

“As a result, many firms have ended up limiting themselves to a small number of search terms to reduce the number of alerts triggered. However, in doing so, they risk missing key signs of market abuse.”

“Our new lexicon completely solves this problem by not only accounting for an unparalleled degree of linguistic variety - but also introducing AI to determine the context of how and where a piece of communication took place. These two fundamental changes enable firms to monitor for a much wider range of risks and greatly reduce the number of false positives.”

Looking ahead, SteelEye plans to further build out more sophisticated capabilities for the product. For example, if an employee typically has phone conversations of more than two minutes and then has a 10 second call, with uncharacteristic intonation - where he or she simply says, “check WhatsApp”, the technology would be able to combine the reference to an unmonitored channel with the unusual call duration to create a more accurate picture of what is going on.

“Lexicons are and will remain important to help firms detect signs of wrongdoing. The industry must focus on getting them right by thinking about searching for language, meaning, and intent rather than individual words or phrases,” concludes Storey.